Research Problems, Questions, Objectives, and Hypotheses in Mechatronics

Introduction

Engineering research begins before an experiment, simulation, prototype, or algorithm is developed. It begins with a defensible decision about what problem matters, why it matters, what is known, what is unknown, and what evidence could resolve the uncertainty. In mechatronics, this decision is especially important because research commonly integrates mechanical structures, sensors, actuators, embedded systems, control, software, communication, and human interaction.

A poorly defined problem can produce technically impressive work with little practical value. Conversely, a well-defined problem gives direction to the literature review, determines the variables to measure, guides the selection of methods, limits the project to a feasible scope, and provides criteria for judging success.

This chapter develops a systematic process for moving from an industry or societal need to a researchable problem, research questions, objectives, hypotheses, assumptions, constraints, and technically feasible deliverables. It also explains how these elements fit together in a research proposal.

Learning objectives

By the end of this chapter, you should be able to:

1. Distinguish a need, problem, research problem, research gap, and research question.
2. Identify and justify an industry or societal need in mechatronics.
3. Define system boundaries, variables, assumptions, and constraints.
4. Formulate clear research questions and aligned research objectives.
5. Construct testable hypotheses and operational definitions.
6. Define success criteria and feasible technical deliverables.
7. Evaluate the feasibility, ethics, safety, and validity of a proposed study.
8. Apply the complete formulation process to a mechatronics case.

1. From need to research problem

1.1 Need, issue, problem, and research problem

A need is a desired improvement or deficiency experienced by a stakeholder. For example, a warehouse may need safer movement of mobile robots around workers. An issue is a broad area of concern, such as robot safety, energy consumption, or agricultural automation. A practical problem is a specific undesirable condition: a mobile robot stops too often near workers, reducing throughput.

A research problem is a clearly bounded uncertainty or deficiency in knowledge, method, design, or evidence that can be investigated systematically. It is not simply a request to build a device. “Design a robot” is a project instruction; “determine whether adaptive speed control reduces near-collision events without unacceptable loss of throughput” is a research problem.

A strong research problem normally contains five elements:

• Context: where and for whom the problem occurs.
• Current condition: what is happening now.
• Desired condition: what improvement is required.
• Knowledge or performance gap: what is not known or not adequate.
• Evidence route: how the uncertainty can be investigated.

A useful formulation is:

In [context], [stakeholder or system] experiences [specific deficiency]. Existing approaches [limitation or uncertainty]. Therefore, research is needed to determine or develop [focused contribution], evaluated using [evidence or criteria].

1.2 Sources of research problems

Research problems may arise from:

• recurring failures in industrial equipment;
• unmet user or societal needs;
• contradictions in published findings;
• limitations of existing sensors, controllers, materials, or algorithms;
• emerging technologies whose performance is not yet established;
• safety, sustainability, accessibility, or regulatory requirements;
• opportunities to transfer a solution from one domain to another;
• observations made during laboratory or field work.

The problem should be supported by evidence. Evidence may include maintenance records, production data, accident reports, user interviews, standards, previous studies, or benchmark tests. A student’s personal interest can motivate a topic, but interest alone does not establish significance.

1.3 Research gap and contribution

A research gap is a justified absence, limitation, inconsistency, or untested assumption in existing knowledge or practice. Common gaps include:

A contribution may be a validated design, experimental evidence, a model, a control strategy, a dataset, a comparative evaluation, or a design method. A contribution need not be a completely new invention. A careful replication in a new operating context can be valuable if the context is important and the study is rigorous.

2. Scope, system boundaries, and feasibility

2.1 Defining scope

Scope states what the study includes and excludes. It should identify the system, operating conditions, users, variables, time period, performance measures, and level of prototype maturity. Without scope, a project expands indefinitely.

For a robotic inspection project, scope might specify:

• a differential-drive indoor robot;
• operation on smooth, level floors;
• speeds between 0.2 and 0.8 m/s;
• one camera and one lidar sensor;
• static obstacles and walking adults;
• evaluation in a laboratory mock-up;
• no claim of certification for unrestricted industrial deployment.

A scope statement is not an admission of weakness. It is a condition for credible conclusions. Results apply only to populations, environments, and operating conditions represented by the study.

2.2 System boundary

A system boundary separates components and interactions being studied from those treated as external. In a mechatronic system, the boundary may include the plant, sensors, controller, actuator drivers, communication network, power supply, and user interface. External influences may include temperature, floor friction, lighting, payload, and human behavior.

A boundary can be represented textually:

Figure: A system boundary distinguishes the components under study from external conditions and measured outputs.

2.3 Assumptions and constraints

An assumption is a condition accepted as sufficiently true for the study, although it may not be directly tested. Examples include assuming that a calibrated encoder remains stable during one trial, or that a battery voltage remains within the controller’s specified range.

A constraint is a limitation imposed by resources, safety, ethics, time, equipment, regulations, or the environment. Examples include a limited sensor budget, a four-month project duration, maximum robot speed in a shared laboratory, or prohibition on collecting identifiable video.

Assumptions should be stated because violated assumptions threaten validity. Constraints should be stated because they determine feasibility. Neither should be hidden in the methods section.

3. Variables and operational definitions

3.1 Types of variables

A variable is a characteristic that can take different values. Typical research variables include:

• Independent variable: the factor deliberately changed or categorized, such as controller type or payload.
• Dependent variable: the measured outcome, such as positioning error or energy consumption.
• Controlled variable: a condition held constant, such as sampling frequency.
• Confounding variable: an uncontrolled factor associated with both the intervention and outcome, such as operator experience.
• Moderating variable: a factor that changes the strength or direction of a relationship, such as floor friction affecting controller performance.
• Mediating variable: an intermediate mechanism through which an intervention influences an outcome.

The same physical quantity can have different roles in different studies. Motor current may be a dependent variable in a study of load estimation and a control input in a study of torque regulation.

3.2 Operational definitions

An operational definition states exactly how a concept will be observed or measured. “Accuracy” is incomplete. It might mean mean absolute position error in millimetres, percentage classification accuracy, or the proportion of successful grasps.

For measured errors, common definitions include:

where y_i is the measured output for trial i, r_i is the reference or target value, and e_i is signed error, all expressed in consistent units.

The mean absolute error for n observations is:

The root mean square error is:

MAE reports the average magnitude of error and is less influenced by large errors than RMSE. RMSE gives greater influence to large deviations because errors are squared. These measures require paired measured and reference values, consistent units, and a clearly defined test set. They are not interchangeable.

3.3 Variable specification table

A variable table exposes ambiguity early and helps determine what instruments, calibration procedures, and analyses are needed.

4. Research questions

A research question converts the problem into an answerable inquiry. It should be specific, neither so broad that it becomes unmanageable nor so narrow that it becomes trivial.

Useful question types include:

• Descriptive: What performance does the current system achieve under specified conditions?
• Comparative: Does controller A perform differently from controller B?
• Relational: How is payload related to positioning error?
• Explanatory: Why does vibration degrade sensor performance?
• Design-oriented: What architecture can satisfy defined requirements?
• Evaluative: To what extent does a prototype meet its acceptance criteria?

A good question identifies the system, relevant conditions, outcome, and comparison or mechanism where appropriate. “How can robots be improved?” is too broad. “Under indoor loads of 0–5 kg, does adaptive velocity control reduce stopping distance compared with fixed-gain control without increasing mean cycle time by more than 10%?” is substantially more researchable.

Research questions should be logically ordered. A design study may ask:

1. What requirements and failure modes characterize the application?
2. What architecture can satisfy the safety and performance requirements?
3. How does the proposed design compare with a baseline?
4. Under what operating conditions does performance deteriorate?

5. Research objectives and success criteria

5.1 Objectives

An objective states what the research will accomplish. Objectives should begin with observable verbs such as identify, characterize, design, develop, implement, compare, estimate, validate, evaluate, or determine. Avoid vague verbs such as “understand” or “study” unless followed by measurable evidence.

The general objective expresses the overall purpose. Specific objectives divide it into achievable outputs or investigations. Each objective should answer four questions:

• What will be done?
• To what system or population?
• Under what conditions?
• How will completion be judged?

Example:

To evaluate whether an adaptive speed controller reduces mobile-robot stopping distance relative to a fixed-speed controller in a laboratory corridor, while maintaining task completion time within a predefined limit.

Specific objectives might be to characterize baseline stopping distance, implement the adaptive controller, measure performance under three obstacle distances, and compare results using predefined metrics.

5.2 Alignment matrix

Misalignment is a common weakness. A proposal may claim to investigate safety but measure only computational speed. The question, objectives, variables, methods, and conclusions must address the same construct.

5.3 Success criteria

Success criteria are observable thresholds or decision rules. They should be defined before testing where practical to reduce biased interpretation. Criteria may concern:

• functional performance;
• accuracy and repeatability;
• response time;
• energy use;
• reliability;
• safety;
• usability;
• cost;
• maintainability;
• environmental impact;
• reproducibility.

A criterion should specify the metric, threshold, test condition, and evidence required. For example: “The prototype shall detect a stationary obstacle at 0.5 m or less with at least 95% detection rate over 100 valid trials.” The criterion is meaningful only if the sensing setup, definition of a valid trial, and test protocol are also specified.

6. Hypotheses and conceptual models

6.1 What a hypothesis is

A hypothesis is a provisional, testable statement about an expected relationship or difference. It must be possible for evidence to contradict it. “The proposed system is better” is vague; “Under identical payload and speed conditions, the proposed controller produces lower RMSE than the baseline controller” is testable.

A directional research hypothesis may be written as:

H₁: The adaptive controller produces a lower mean stopping distance than the fixed-speed controller under the specified test conditions.

A null hypothesis expresses no difference or relationship:

H₀: The adaptive and fixed-speed controllers have equal mean stopping distance under the specified test conditions.

Not every project requires a formal statistical hypothesis. Exploratory, qualitative, and early design research may use research questions and evaluation criteria instead. A hypothesis is appropriate when theory or prior evidence supports a prediction and the variables can be measured reliably.

6.2 Hypothesis quality

A useful hypothesis is:

• conceptually justified;
• stated in terms of observable variables;
• directional when a direction is justified;
• feasible to test with available resources;
• limited to the population and conditions studied;
• compatible with the proposed design.

Do not formulate a hypothesis after inspecting results merely to make the outcome appear predicted. Distinguish pre-specified hypotheses from post hoc explanations.

6.3 Conceptual framework

A conceptual framework maps the expected relationship among inputs, mechanisms, outcomes, and conditions. For example:

Figure: A conceptual framework identifies the proposed mechanism, outcomes, and conditions that may influence results.

The framework is not proof of causation. It is a transparent statement of the logic that the study will examine.

7. Technical feasibility and deliverables

A technically feasible project matches ambition to time, competence, equipment, budget, access, and safety. Feasibility should be assessed before objectives are finalized.

Important feasibility questions include:

• Are the required sensors, actuators, software, and computing resources available?
• Can the independent variable be manipulated safely and consistently?
• Can the dependent variables be measured with adequate resolution and calibration?
• Is a baseline or comparison condition available?
• Can enough valid trials be completed?
• Are the necessary users, datasets, or industrial records accessible?
• Can the project be completed within the academic calendar?
• What happens if a key component fails or data access is denied?

Deliverables should be distinguishable from activities. “Conduct testing” is an activity; “calibrated dataset and evaluation report” is a deliverable. A realistic undergraduate mechatronics project may deliver:

1. requirements and risk specification;
2. system architecture and interface definitions;
3. simulation model or digital prototype;
4. functioning laboratory prototype;
5. source code with version history;
6. calibration and test protocol;
7. curated dataset;
8. performance evaluation against a baseline;
9. safety and limitations statement;
10. reproducible technical report.

A deliverable is not automatically a research contribution. The contribution lies in the justified knowledge or evidence produced through the design and evaluation.

8. Worked formulation example

Consider a small manufacturing cell in which a vision-guided pick-and-place robot occasionally misplaces reflective components. Production staff report rework, but the causes are uncertain.

Initial need: Reduce rework caused by inaccurate placement.

Evidence review: Preliminary records show more errors under changing illumination and for highly reflective parts. The existing camera uses fixed exposure and a conventional thresholding method.

Research problem: The cell lacks evidence about whether adaptive image preprocessing improves placement accuracy across realistic illumination conditions, and the effect on cycle time is unknown.

Scope: One six-axis laboratory robot, one camera, three component types, three controlled illumination levels, and a fixed placement surface. The study excludes production certification and long-term maintenance reliability.

Variables:

• independent variable: image-processing method, conventional or adaptive;
• environmental variable: illumination level, measured in lux;
• dependent variables: placement error in millimetres, detection success rate, and cycle time in seconds;
• controlled variables: robot program, camera location, component feeder, and target coordinates.

Research question: Under three defined illumination levels, does adaptive image preprocessing reduce mean placement error compared with conventional thresholding without increasing mean cycle time by more than 10%?

General objective: To evaluate adaptive image preprocessing for improving vision-guided placement of reflective components.

Hypothesis: Under identical robot and camera conditions, adaptive preprocessing produces lower mean placement error than conventional thresholding.

Success criteria: The adaptive method must reduce MAE by at least 20%, maintain a detection success rate of at least 95%, and increase mean cycle time by no more than 10% relative to the baseline. These thresholds are engineering decisions that must be justified by stakeholder needs, prior evidence, or system requirements.

Deliverables: documented requirements, two implemented algorithms, calibrated test setup, raw and processed dataset, statistical comparison, source code, and limitations report.

The formulation is stronger than “improve robot vision” because it defines the context, comparison, variables, operating conditions, metrics, and decision rules.

9. Common weaknesses and quality checks

9.1 Common weaknesses

• Topic instead of problem: “Artificial intelligence in robotics” identifies an area, not an uncertainty.
• Overly broad scope: attempting to address sensing, mechanics, control, cloud computing, and commercialization in one undergraduate project.
• Unmeasurable objectives: using “make the system intelligent” without operational criteria.
• Solution bias: deciding on a technology before confirming that it addresses the need.
• No baseline: reporting prototype performance without comparison.
• Hidden assumptions: treating ideal lighting, perfect communication, or constant loads as universal conditions.
• Metric mismatch: claiming improved safety while measuring only accuracy.
• Unrealistic deliverables: promising industrial deployment when only a laboratory prototype can be validated.
• Hypothesis after the fact: selecting predictions after observing data.

9.2 Formulation checklist

Before approving a proposal, ask:

1. Who experiences the need, and what evidence demonstrates it?
2. What precise uncertainty or deficiency is being investigated?
3. What is inside and outside the system boundary?
4. Are all important variables operationally defined?
5. Does each objective produce an observable result?
6. Can each research question be answered by the proposed evidence?
7. Is a hypothesis justified, and can it be contradicted?
8. Are success criteria specified before evaluation?
9. Are safety, privacy, intellectual property, and regulatory issues addressed?
10. Can the deliverables be completed with available resources?

10. Ethical, safety, and governance considerations

Mechatronics research can affect people physically, socially, and economically. A prototype robot, exoskeleton, autonomous vehicle, or medical device must be assessed for foreseeable hazards. Risk controls should include safe operating modes, emergency stops, guards, current or force limits, supervised testing, and documented failure procedures.

If cameras, wearable sensors, location data, or user performance data are collected, researchers must address informed consent, data minimization, secure storage, access control, anonymization where possible, and approved retention periods. Industrial data may contain confidential process information or intellectual property.

AI-enabled systems introduce additional concerns. A model may perform well on laboratory data but fail under unrepresented lighting, users, loads, or environmental conditions. Research questions should therefore address data provenance, test-set separation, generalization, explainability where relevant, and human oversight. Claims should not exceed the tested operating domain.

Standards and regulations may define requirements for machinery safety, electrical safety, electromagnetic compatibility, functional safety, medical devices, or data protection. Compliance cannot be claimed merely because a prototype passes a student-designed test. Where certification is outside the project scope, that limitation must be explicit.

Contemporary developments and professional relevance

Modern mechatronics research increasingly uses digital twins, edge computing, collaborative robots, additive manufacturing, machine learning, and cyber-physical systems. These technologies expand the possible solution space but also make problem formulation more important. A digital twin may support simulation, but the research question must specify whether the issue concerns model fidelity, predictive accuracy, or decision usefulness. Machine learning may improve classification, but the study must define data conditions, error costs, and generalization requirements.

In professional practice, engineers are often evaluated not only on whether they can build a system, but on whether they can define the right problem, justify requirements, manage risk, communicate uncertainty, and produce evidence that supports a decision. The ability to align need, question, objective, measurement, and deliverable is therefore a core employability skill.

Chapter Summary

A research project begins with a significant and evidenced need, but a need must be transformed into a bounded research problem. The research problem identifies a specific deficiency or uncertainty in a defined context and points toward an evidence-based contribution. Scope and system boundaries prevent uncontrolled expansion; assumptions and constraints make the conditions of validity visible.

Variables must be classified and operationally defined so that abstract concepts such as accuracy, safety, or efficiency become measurable. Research questions should be focused and answerable. Objectives convert those questions into observable actions and outputs. Hypotheses state testable predictions when theory and measurement justify them, while exploratory studies may rely on questions and evaluation criteria instead. Success criteria establish how achievement will be judged, and feasible deliverables connect research activity to professional practice.

Strong proposals align need, problem, scope, variables, questions, objectives, hypotheses, methods, criteria, and deliverables. Ethical, safety, privacy, intellectual-property, and regulatory considerations must be incorporated from the beginning rather than added after technical development.

Key Terms

Research problem: A bounded uncertainty or deficiency that can be investigated systematically.

Research gap: An absence, limitation, inconsistency, or untested condition in existing knowledge or practice.

Scope: The stated limits of a study.

System boundary: The distinction between elements included in the investigation and external influences.

Assumption: A condition accepted as sufficiently true for the study.

Constraint: A limitation imposed by resources, safety, ethics, time, or regulations.

Variable: A characteristic that can take different values.

Independent variable: A factor changed or compared by the researcher.

Dependent variable: An outcome measured in the study.

Controlled variable: A condition held constant.

Confounding variable: An uncontrolled factor that can distort an observed relationship.

Operational definition: A precise description of how a concept is measured or implemented.

Research question: An answerable inquiry derived from the research problem.

Research objective: An observable accomplishment pursued by the study.

Hypothesis: A testable prediction about a relationship or difference.

Null hypothesis: A statement that a specified difference or relationship is absent.

Success criterion: A predefined metric, threshold, and test condition used to judge achievement.

Baseline: A reference method or condition used for comparison.

Deliverable: A tangible output produced by the project.

Conceptual framework: A representation of expected relationships among variables and mechanisms.

Review Questions

Short Answer Questions

1. Distinguish a practical problem from a research problem.
2. Why is an operational definition necessary for a variable such as accuracy?
3. What is the difference between an assumption and a constraint?
4. State two characteristics of a well-formed research objective.
5. When might a research project use questions and success criteria without a formal hypothesis?

Long Answer Questions

1. Explain how an industry need can be transformed into a research problem, research question, objectives, and deliverables.
2. Discuss the importance of defining a system boundary in a mechatronics research project.
3. Compare descriptive, comparative, relational, explanatory, and evaluative research questions, using mechatronics examples.
4. Explain how hidden assumptions and uncontrolled variables can threaten the validity of an engineering study.
5. Discuss ethical, safety, privacy, and regulatory issues that should be considered when researching an AI-enabled robot.

Analytical / Application Questions

1. A conveyor inspection system has high detection accuracy but rejects too many acceptable products. Formulate a research problem, one research question, two objectives, variables, and two success criteria.
2. A student proposes to “develop an autonomous agricultural robot” in one semester. Critically evaluate the scope and rewrite it as a technically feasible undergraduate research project.
3. Design an alignment matrix for a wearable rehabilitation device, including the need, problem, question, objective, independent variable, dependent variables, hypothesis, baseline, and deliverables.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a research problem?

A. A broad topic such as robotics

B. A request to purchase laboratory equipment

C. A bounded uncertainty or deficiency that can be investigated systematically

D. A list of all technologies used in a system

Correct Answer: C. A bounded uncertainty or deficiency that can be investigated systematically

Question 2

Which is an operational definition of robot positioning accuracy?

A. The robot is highly precise

B. The robot uses an advanced controller

C. The average absolute difference between target and measured position in millimetres

D. The robot is suitable for industry

Correct Answer: C. The average absolute difference between target and measured position in millimetres

Question 3

In a controller comparison, controller type is usually the:

A. Dependent variable

B. Independent variable

C. Confounding variable

D. Measurement error

Correct Answer: B. Independent variable

Question 4

Which objective uses the most observable verb?

A. Understand the robot better

B. Study automation generally

C. Evaluate positioning error under three payload conditions

D. Appreciate the benefits of sensing

Correct Answer: C. Evaluate positioning error under three payload conditions

Question 5

What is the main purpose of a baseline?

A. To eliminate the need for measurements

B. To provide a reference condition for comparison

C. To guarantee that a hypothesis is true

D. To replace safety testing

Correct Answer: B. To provide a reference condition for comparison

Question 6

Which statement is a testable hypothesis?

A. The proposed robot is excellent

B. Automation is important in society

C. Adaptive control produces lower stopping distance than fixed-speed control under defined conditions

D. Robots should be improved

Correct Answer: C. Adaptive control produces lower stopping distance than fixed-speed control under defined conditions

Question 7

Which item is a constraint rather than an assumption?

A. The encoder remains calibrated during one trial

B. The laboratory floor is level

C. Testing must be completed within a four-month project period

D. The battery voltage remains within its rated range

Correct Answer: C. Testing must be completed within a four-month project period

Question 8

For paired errors eᵢ, which metric gives greater influence to large errors?

A. RMSE

B. Median only

C. The number of controlled variables

D. The system boundary

Correct Answer: A. RMSE

Question 9

Which feature most clearly indicates an over-broad project scope?

A. A defined prototype and test environment

B. Specified variables and success criteria

C. An attempt to address mechanics, control, cloud computing, certification, and commercialization in one semester

D. A comparison with a baseline

Correct Answer: C. An attempt to address mechanics, control, cloud computing, certification, and commercialization in one semester

Question 10

Which is an appropriate success criterion?

A. The system shall be intelligent

B. The prototype shall feel modern

C. The detector shall achieve at least 95% detection rate over 100 valid trials

D. The researcher shall improve the algorithm

Correct Answer: C. The detector shall achieve at least 95% detection rate over 100 valid trials